Gym入门使用教程

The Gym interface is simple, pythonic, and capable of representing general RL problems:

import gym
env = gym.make("LunarLander-v2")
observation, info = env.reset(seed=42)
for _ in range(1000):
   env.render()
   action = policy(observation)  # User-defined policy function
   observation, reward, done, info = env.step(action)

   if done:
      observation, info = env.reset()
env.close()

基础

官方文档

一，激活environment,查看环境基本信息

env.observation_space 得到state信息，是一个Box类，
env.observation_space.shape 得到state的shape
env.action_space 得到action的信息，是一个Discrete类
env.action_space.n 得到action的个数

import gym,time
import random
import numpy as np
#初始化环境这里选择三个不同类别的环境
env1 = gym.make('LunarLander-v2')
env2 = gym.make('Pong-v0')
env3 = gym.make('CartPole-v0')
#查看环境状态
#可以看到观察环境空间状态信息，主要是环境相关矩阵，一般是一个box类
print(env1.observation_space, type(env1.observation_space))
print(env2.observation_space, type(env2.observation_space))
print(env3.observation_space, type(env3.observation_space))
#返回的是离散action空间的值
print(env1.action_space, type(env1.action_space))
print(env2.action_space, type(env2.action_space))
print(env3.action_space, type(env3.action_space))
#LunarLander-v2的state shape和action space大小
print(env1.observation_space.shape, env1.action_space.n)
#Pong-v0的state shape和action space大小
print(env2.observation_space.shape, env2.action_space.n)

二，使用reset初始化environment,查看state信息（换个游戏场景）

state1 = env1.reset()
state2 = env2.reset()
state3 =env3.reset()

三，执行action并使用render可视化

这里主要使用env.setp来执行，输入值为一个action的序号。返回值为new state,action reward,action terminal bool 和一个其他信息

state = env1.reset()
env1.render()
new_state, reward, done, info = env1.step(0)
print(reward, done, info)

四，如何执行完一个episodic

env = gym.make('CartPole-v0')
for i_episode in range(20):
    observation = env.reset() #初始化环境每次迭代
    for t in range(100):
        env.render() #显示
        print(observation)
        action = env.action_space.sample() #随机选择action
        observation, reward, done, info = env.step(action)
        if done:
            print("Episode finished after {} timesteps".format(t+1))
            break
env.close()

环境

自定义环境

import gym

%matplotlib inline
from matplotlib import pyplot as plt
#gym.version==0.26


#定义环境
class MyWrapper(gym.Wrapper):

    def __init__(self):
        env = gym.make('CartPole-v1', render_mode='rgb_array')
        super().__init__(env)
        self.env = env
        self.step_n = 0

    def reset(self):
        state, _ = self.env.reset()
        self.step_n = 0
        return state

    def step(self, action):
        state, reward, done, _, info = self.env.step(action)

        #一局游戏最多走N步
        self.step_n += 1
        if self.step_n >= 200:
            done = True

        return state, reward, done, info
    
env = MyWrapper()
env.reset()

#认识游戏环境
def test_env():
    print('env.observation_space=', env.observation_space)
    print('env.action_space=', env.action_space)

    state = env.reset()
    action = env.action_space.sample()
    next_state, reward, done, _ = env.step(action)

    print('state=', state)
    print('action=', action)
    print('next_state=', next_state)
    print('reward=', reward)
    print('done=', done)


test_env()



#打印游戏
def show():
    plt.figure(figsize=(3, 3))
    plt.imshow(env.render())
    plt.show()


show()

Classic Control

Cart Pole

Action Space	Discrete(2)
Observation Shape	(4,)
Observation High	[4.8 inf 0.42 inf]
Observation Low	[-4.8 -inf -0.42 -inf]
Import	gym.make("CartPole-v1")

Mountain Car Continuous

Action Space	Box(-1.0, 1.0, (1,), float32)
Observation Shape	(2,)
Observation High	[0.6 0.07]
Observation Low	[-1.2 -0.07]
Import	gym.make("MountainCarContinuous-v0")